It has been hypothesized that females mate multiply to increase the heterozygosity of their progeny because heterozygous individuals are assumed to have a fitness advantage. Females can maximize heterozygosity of their offspring by mating with genetically unrelated and/or heterozygous males. We tested these predictions in a socially monogamous passerine, the reed bunting ( Emberiza schoeniclus ), where extrapair paternity occurs frequently. The results based on genotypes at nine microsatellite loci revealed that females were no less genetically related to the extrapair male(s) (EPMs) than to their pair male (i.e., breeding partner) and that EPMs were no more heterozygous than the males they cuckolded. In addition, a direct comparison of maternal half-siblings naturally raised in the same brood showed that extrapair young were no more heterozygous than within-pair young. Thus, female reed buntings do not seem to mate with EPMs to increase offspring heterozygosity. It is not yet known whether extrapair mating involves any benefits at all to females in this species.
Extrapair paternity (EPP) is present in the majority of the bird species for which paternity has been investigated (reviewed in Griffith et al., 2002 ). In the last two decades extensive effort has been made to obtain a thorough insight into the patterns and mechanisms of EPP. However, the adaptive value of extrapair fertilizations (EPFs) is still poorly understood and highly debated, especially so for females (e.g., Griffiths et al., 1998 ; Jennions and Petrie, 2000 ; Tregenza and Wedell, 2000 ; Westneat and Stewart, 2003 ; Zeh JA and Zeh DW, 2003 ). In the absence of direct benefits, several hypotheses have been put forward in an attempt to explain the possible genetic (indirect) advantages that females may gain by mating with males outside their pair bond (reviewed in Jennions and Petrie, 2000 ).
Previous studies have shown that the individual level of inbreeding (measured as heterozygosity) can affect an individual's survival and fitness (e.g., Allendorf and Leary, 1986 ; Coltman et al., 1998 ; Foerster et al., 2003 ; Hansson et al., 2001 ; Keller and Waller, 2002 ; Slate et al., 2000 , but see e.g., David, 1998 ; Whitlock, 1993 ). It has been proposed that females mate multiply to maximize offspring heterozygosity (e.g., Brown, 1997 ) and thereby fitness. Females may enhance the heterozygosity of their progeny by mating with males who are either less genetically related to themselves or by mating with males who are heterozygous ( Brown, 1997 ; Mitton et al., 1993 ). Because females in socially monogamous species may be constrained in their mate choice ( Møller, 1994 ; Petrie and Kempenaers, 1998 ) they might end up with a genetically suboptimal male. Hence, it has been suggested that EPFs could be a way for females to circumvent this restriction on mate choice and thereby increase the genetic quality (e.g., heterozygosity) and the fitness of their offspring ( Trivers, 1972 ).
Testing the heterozygosity hypothesis by investigating whether females use extrapair copulations (EPCs) to enhance offspring heterozygosity has rarely been performed. Until recently, the few attempts that have been made have used an indirect approach, for instance by looking at female extrapair mate preference (e.g., Masters et al., 2003 ). To our knowledge there is only one study that has tested the hypothesis in a design where maternal half-siblings sired by different males have been compared. Foerster et al. (2003) recently showed that female blue tits ( Parus caruleus ) increased offspring heterozygosity through EPFs, thereby supporting the heterozygosity hypothesis.
The aim of our study was to repeat Foerster et al.'s (2003) approach in another socially monogamous passerine with an extensive extrapair mating system, the reed bunting ( Emberiza schoeniclu s). Its level of EPP is among the highest reported for a bird species (reviewed in Griffith et al., 2002 ). If female reed buntings engage in EPCs to increase the heterozygosity of their offspring, we predict (1) the females to be less genetically similar to the extrapair male (EPM) than to their pair male (i.e., breeding partner); (2) and/or the EPM(s) to be more heterozygous than the pair male; consequently, we predict (3) the extrapair young to be more heterozygous than within-pair young (WPY) in the same brood.
Study area and fieldwork
Reed buntings were studied in the valley Øvre Heimdalen (61° 25′ N, 8° 52′ E), in southern Norway, during the breeding seasons 2001 and 2002. General field methods are described in detail elsewhere ( Kleven and Lifjeld, 2004 ) and only briefly outlined here. Adults were caught with mist nets, and we provided each individual with a unique combination of a numbered aluminum ring and three color rings for later identification. At the same time, a small amount of blood was sampled by puncturing the brachial vein. Two days after, hatching nestlings were bled (3–25 μl) by puncturing either the brachial or the femoral vein.
We used nine microsatellite loci to determine paternity of nestlings as well as to estimate individual heterozygosity and relatedness among pairs of individuals. A detailed description of the genotyping methods is presented in Kleven and Lifjeld (2004) . Briefly, we optimized and used microsatellite primers from the reed bunting and five other bird species (Ase18: Richardson et al., 2000 ; Cuμ4: Gibbs et al., 1999 ; Escμ1, Escμ3, Escμ4 and Escμ6: Hanotte et al., 1994 ; Mcyμ4: Double et al., 1997 ; Pdoμ5: Griffith et al., 1999 and Ppi2: Martinez et al., 1999 ). Nestlings were defined as WPY if they matched (i.e., possessed alleles compatible with Mendelian inheritance from) their social parents completely ( n = 234). The probability of chance inclusion ( Jeffreys et al., 1992 ) was found to be low (range; 1.1 × 10 −9 to 9.7 × 10 −4 ). A mismatch with the genotype of a putative parent can either arise from mutation or mismatched parentage. Mutation rates at microsatellite loci are usually in the order of 10 −3 or lower per meiotic event (reviewed in Ellegren, 2000 ), which means that mutations would normally be rare. However, to account for the probability of one mismatch due to a mutation, we defined young as extrapair offspring if they did not match the putative father at two or more loci. In our sample, there were no offspring with a genotype mismatch at only a single locus. All 332 offspring had a full allelic match with the putative mother, whereas 98 offspring had alleles that did not match the genotype of the putative father at two or more loci. We thus consider the latter group to be sired by EPMs. Paternity of extrapair young (EPY) was assigned to a male if there was a match at all loci. All genotyped males were screened for the paternally inherited alleles found in EPY. The combined probability of paternal exclusion ( Jamieson, 1994 ) for all nine loci was higher than 0.999. The observed heterozygosity at the nine microsatellite loci ranged from 0.52 to 0.93 ( Kleven and Lifjeld, 2004 ). We calculated standardized individual heterozygosity (proportion of heterozygous loci/mean heterozygosity of typed loci; Coltman et al., 1999 ) because a few individuals did not amplify alleles at all loci. Hereafter, we refer to this measure as heterozygosity. As a measure of genetic similarity, we calculated pairwise genetic relatedness (see Queller and Goodnight, 1989 ) using the software program Relatedness 5.0.8 ( http://www.gsoftnet.us/GSoft.html ). We also performed the analyses with another measure of genetic diversity, d2 ( Coulson et al., 1998 ). The results using d2 were essentially similar to those using heterozygosity and are therefore not included. In clutches with more than one EPM siring offspring, the female's relatedness to the sires and the heterozygosity of the sires were averaged.
We used a restricted maximum likelihood (REML) design to test for differences in heterozygosity between maternal half-siblings (WPY and EPY) raised in the same brood because of the unbalanced structure of the data. Nest was defined as a random factor in the analysis. To examine the number of EPY in each brood in relation to heterozygosity, we used generalized linear models (GLMs) with binomial errors and logit link. Brood size was used as a binomial denominator. Number of EPY sired and total fertilization success for a male in relation to heterozygosity were analyzed with GLMs with Poisson error distribution and logarithm link. Statistical analyses were performed using SPSS (version 11.0), STATISTICA (version 6.1), and GLMstat (version 5.5). Confidence intervals (95%) are given for the main statistical tests ( Colegrave and Ruxton, 2003 ).
Several of the adults (16/44 males and 8/35 females) that were caught in the study area during the first field season returned in the second year. Breeding attempts in which the male or female changed partners between the two seasons were treated as independent attempts. Each individual was, however, only included once when testing for heterozygosity effects in relation to paternity. In these cases, we used the first breeding attempt in which data on brood paternity was obtained. Hence, the sample sizes in the statistical analyses vary accordingly. All tests were two-tailed, and the assumed level of significance for all tests was p < .05.
EPP was common in our reed bunting population. Combined for the 2 years, 54% (39/72) of all broods contained at least one EPY, and 30% (98/332) of the nestlings were sired by an EPM. We were able to assign paternity to 61% (60/98) of the EPY. In three broods, we only knew the identity of the mother and so were unable to determine the paternity of the nestlings.
Heterozygosity, relatedness, and within-pair paternity
Pair formation did not seem to be influenced by heterozygosity as males and females did not pair up assortatively based on heterozygosity ( r = −.04, p = .77, n = 68). We found that the likelihood of being cuckolded could not be predicted by a male's heterozygosity or the relatedness to the female he was paired up with (binomial logistic regression; heterozygosity: χ 2 = 1.62, df = 1, p = .20, n = 53; relatedness: χ 2 = 1.24, df = 1, p = .27, n = 53). Heterozygosity and relatedness to the female also failed to predict the proportion of young a male sired in his own brood (GLM; heterozygosity: χ 2 = 0.62, df = 1, p = .54, n = 53; relatedness: χ 2 = 1.14, df = 1, p = .26, n = 53). Female heterozygosity showed a weak but nonsignificant tendency to predict the likelihood of having one or more eggs fertilized by EPMs (binomial logistic regression; χ 2 = 3.22, df = 1, p = .07, n = 62) and failed to predict the proportion of young fertilized by an EPM(s) (GLM; χ 2 = 0.86, df = 1, p = .39, n = 59).
Heterozygosity and EPP
The heterozygosity of a male neither predicted the probability of him fertilizing eggs outside his own nest (logistic regression; χ 2 = 0.50, df = 1, p = .48, n = 68) nor predicted the number of EPFs he obtained (GLM; χ 2 = 1.89, df = 1, p = .17, n = 68).
Comparison of cuckoldee and cuckolder
A pairwise comparison of the pair male and the cuckolder revealed that there were no significant differences between the two groups of males with respect to both heterozygosity ( t19 = −0.58, p = .57; 95% confidence interval: −0.18 to 0.10; Figure 1a ) and relatedness to the female ( t19 = 0.44, p = .67; 95% confidence interval: −0.10 to 0.15; Figure 1b ). These results were not in accordance with our predictions 1 and 2, respectively. As a further test of extrapair mating preferences, the closest neighbor to a focal female that did not obtain paternity was also compared with the actual cuckolder. The two groups of males did not differ significantly from each other in heterozygosity or the relatedness to the female they were paired up with (paired t test; heterozygosity: t19 = −0.78, p = .44; relatedness: t19 = 1.21, p = .24).
EPP and offspring heterozygosity
In total, 30 out of 69 paternally assigned broods contained young with mixed paternity (76 WPY, and 71 EPY) for which heterozygosity could be compared. Contrary to prediction 3, we found that there were no significant difference in heterozygosity between WPY and EPY (REML; F1,116 = 0.09, p = .77; 95% confidence interval: −0.04, 0.06; Figure 2 ). In their blue tit study, Foerster et al. (2003) found that the difference in heterozygosity among maternal half-siblings was mainly due to females mating with nonlocal males. To evaluate if the same pattern could be present in our reed bunting population, we compared heterozygosity of maternal half-siblings in nests for which we knew the identity of the EPM (defined as local males) and in nests for which we did not have the identity of the EPM (assumed to be nonlocal males) separately. In this comparison one brood was included in both groups as it contained young with an identified and an unidentified EPM. There was no significant difference between maternal half-siblings in either the former (REML; F1,65 = 0.46, p = .50, n = 17 broods) or the latter group (REML; F1,54 = 0.67, p = .42, n = 14 broods).
Heterozygosity and fitness
Neither male nor female heterozygosity was related to the date of breeding onset ( rs = .12, p = .35, n = 61; rs = 0.17, p = .18, n = 64; respectively). Female heterozygosity was also not correlated with clutch size ( rs = −.12, p = .41, n = 49). Females may mate multiply to improve egg hatchability ( Kempenaers et al., 1999 ). Hatching success (no. of hatched eggs/no. of eggs laid) in a brood was, however, similar whether it contained EPY (mean ± SD; 0.89 ± 0.19, n = 25 broods) or not (mean ± SD; 0.90 ± 0.19, n = 21 broods; ANOVA: F1,45 = 0.02, p = .89). Heterozygosity of a male was not a predictor of his total fertilization success (GLM; χ 2 = 1.44, df = 1, p = .30, n = 56), although it should be noted that males may have obtained EPFs in broods outside our study area. The relatedness between the male and female in a pair was not significantly correlated with clutch size (partial correlation controlling for breeding date; r = −.10, p = .47, n = 54 broods) or hatching success ( rs = −.19, p = .22, n = 45 broods).
The hypothesis that females engage in EPCs to increase offspring heterozygosity was not supported in this study of reed buntings. Theoretically, females can maximize heterozygosity of their progeny by mating with either heterozygous or genetically dissimilar males. Mating with individuals that reduce the probability of inbreeding is assumed to be advantageous because inbreeding has been shown to be associated with decreased fitness arising through the expression of recessive deleterious alleles due to homozygosity (reviewed in Keller and Waller, 2002 ). We found that EPMs were no more heterozygous nor less genetically related to the female than her pair mate. Also, extrapair offspring were no more heterozygous than their maternal half-siblings raised in the same brood.
Females may mate multiply to ensure that they mate with a genetically compatible sire ( Tregenza and Wedell, 2000 ; Zeh JA and Zeh DW, 1997 ). Although we did not find that female reed buntings mated with EPMs to increase offspring heterozygosity at a number of microsatellite loci (this study) or to enhance offspring immunocompetence ( Kleven and Lifjeld, 2004 ), we cannot exclude the possibility that they seek mates that are more compatible at other loci. For example, the major histocompatibility complex genes may be likely candidates as they are associated with fitness in other species (reviewed in Jennions and Petrie, 2000 ; Penn and Potts, 1999 ).
The present lack of a relationship between heterozygosity and EPP may have other possible explanations. The hypothesis that EPP is a way for females to increase offspring heterozygosity has two key assumptions. First, individual heterozygosity has to be associated with fitness. In our reed bunting population, we found no indication that this was the case. It must be pointed out, however, that the fitness measures used relate to reproduction and not viability. The assumption of heterozygosity benefits may, however, be valid only in populations with a certain genetic structure (e.g., locally inbred or recently bottlenecked populations) and not for large panmictic and outbred populations ( David, 1998 ; Hansson and Westerberg, 2002 ). In the blue tit population studied by Foerster et al. (2003) a spatial genetic structuring was documented and individual heterozygosity was also linked to fitness. Similarly, a study on a recently founded population of great reed warblers ( Acrocephalus arundinaceus ) also showed an association between individual heterozygosity and fitness ( Hansson et al., 2001 ). We have only indirect information regarding the presence or absence of a genetic structuring in our population of reed buntings. A genetic structure in our population could be indicated if the difference in heterozygosity between maternal half-siblings was mainly related to whether females mated extrapair with local or nonlocal males, as shown by Foerster et al. (2003) . However, we failed to find such an effect. If the reed bunting population studied represents a large panmictic population without genetic structuring, females may therefore not be expected to choose EPMs to improve offspring heterozygosity.
The second assumption from the heterozygosity hypothesis is that females pursue EPCs. There is, however, great controversy over this issue ( Westneat and Stewart, 2003 ). It is not yet known which sex controls EPCs in the reed bunting, but behavioral data indicate that only males perform extraterritorial forays ( Marthinsen et al., in press ). Moreover there seems to be no direct (Kleven O and Marthinsen G, unpublished observations) or indirect ( Kleven and Lifjeld, 2004 ; this study) benefits involved for females in extrapair mating. In another population of this species there is some evidence of direct costs (reduced paternal care) to females involved in EPFs ( Dixon et al., 1994 ). The reed bunting mating system may therefore be driven by male rather than female benefits, and female cooperation may be selected through the reduction of resistance costs ( Westneat and Stewart, 2003 ). If so, increased offspring heterozygosity resulting from EPFs may not be expected.
Another potential problem in interpreting our results is that the measure of heterozygosity was based on a relatively small number ( n = 9) of microsatellite loci. The power to detect a link between heterozygosity and fitness, and to detect a difference in heterozygosity between maternal half-siblings, could therefore be low, as indicated by Slate and Pemberton (2002) . In addition, the results presented by Foerster et al. (2003) revealed that the effects of female extrapair mating on offspring heterozygosity were not very strong as they found that the mean difference in heterozygosity between WPY and EPY was only 0.02 (equivalent to 2.0%). Being able to detect a potential effect of a female preference for EPMs that are either more heterozygous or less related than the pair male thus probably requires a combination of a large number of microsatellite markers and a large sample size. With our sample size we should have been able to detect a mean difference in heterozygosity of 0.08 (equivalent to 7.7%) between the two groups when setting the power to 0.8 and α to 0.05. Based on this we would have found a significant difference between the maternal half-siblings if EPY had showed a mean heterozygosity value of 1.12 (mean value observed for WPY, i.e., 1.04, plus 0.08). Therefore, our sample size did not allow us to detect a comparably weak effect as that reported in the study by Foerster et al. (2003) . However, our data did not show any strong tendencies in support of EPY being more heterozygous than WPY, that is, the mean heterozygosity value was lower in EPY compared to WPY (see Figure 2 ), and EPY were more heterozygous than WPY in only half of the broods (see Figure 2 ).
In summary, our results suggest that females in this population of reed buntings did not obtain EPFs with males that were less genetically similar to them or more heterozygous than their pair mate. We also found no indication that females obtained indirect (genetic) benefits, in terms of increased offspring heterozygosity, from EPFs. The hypothesis that females mate multiply to increase offspring heterozygosity is therefore not supported.
We thank Frode Fossøy, Jon B. Kirkebø, Roy Mangersnes, Henrik Pärn, Roar Rismark, and Dave Showler for their help and support during the fieldwork. Erik Brenna, Frode Fossøy, Lars Erik Johannessen, Arild Johnsen, and Mary Stapleton gave valuable comments on an earlier version of the manuscript. We are especially indebted to Erik Brenna and Gunnhild Marthinsen for their help both in the field and with the lab work. The study was supported by a PhD grant (O.K.) from the Research Council of Norway, and it was approved by the Norwegian Animal Research Authority (License no. S-1236/01).